This article is by Phil Worthey, see bio below;
There is a particular kind of delay that still catches claims handlers out. A motor or public liability file has been under assessment for some time when someone finally notices that the same knee, shoulder or lumbar spine was treated two years before the incident, or that documented care stopped for several months and then restarted. The fact was often already in the production, and simply late to be seen.
If a prior condition or a quiet stretch in treatment appears only after a view of the claim has begun to settle, the team has to retrace its steps. The useful question is whether organising the file earlier can bring those issues into view sooner, and what a claims handler should still do once they are on the screen.
Why those facts arrive late
Medical records rarely arrive as a single, tidy history. They come from A&E departments, GPs, specialists, imaging centres and physiotherapy clinics, often as a mix of PDFs, scans and forwarded packets. If a handler works through that pile in the order it arrived, earlier treatment and a later follow-up can sit many pages apart. A quiet interval in care is easy to miss, and duplicates make the course of treatment look busier than it was.
This is the part of the work AI is well suited to. It identifies and removes duplicate and near-duplicate pages, groups the remainder by provider and by the date treatment actually took place, and drafts a chronology in which each entry cites its source page. Done properly, that can be available in hours. The point is a file in which a prior attendance for the same body part, and a gap between attendances, are visible near the start of the assessment rather than near the end.
What earlier sight is worth
A source-cited chronology does not tell an insurer what to pay. It changes the order in which a handler can ask useful questions. If earlier treatment of the same region is on the timeline, and each row points to a page, the team can read that history against the post-incident notes instead of discovering it later. If a period with no documented care is flagged, the team can ask whether records are still outstanding, or whether treatment simply stopped.
Fictional illustration.
This example is invented. A handler reviewing a sorted chronology sees a GP attendance for right-shoulder pain in March 2023, a workplace incident involving the same shoulder in January 2025, and no documented physiotherapy between a February 2025 referral and a June 2025 review, each cited to a page. The chronology has not decided the claim. It has put those facts where the handler can check them in time. Work of this kind, including the source-cited chronologies described for insurance claims teams, is only as useful as the citation behind each entry.
What still belongs to the claims handler
Confirm the pages you will rely on. If a prior condition, a first attendance after the incident, a procedure or a treatment gap will influence the assessment, open the cited page and confirm that the date, the provider and the wording are actually there. If the page does not support the entry, treat the entry as unproven.
Treat flags as questions, not findings. A gap is not proof that the claimant stopped treating, and it is not proof that a record is missing. The notes may refer to a clinic whose records never arrived, or they may explain a planned break in care. The handler decides what the interval means.
Do not let an organised file stand in for a complete one. If a referral, an imaging report or a consultant letter is mentioned and not present, the chronology can only describe the production you have. Sorting that production does not show you what was never sent.
Keep causation where it belongs. A timeline can show that earlier treatment exists, or that symptoms were recorded after an incident. It cannot prove that the incident caused the injury, or that it did not. That judgement sits with the claims team, with insurance-defence counsel where they are instructed, and, where the medical question is material, with an independent medical expert. AI output is not a medical opinion, and it is not a substitute for one.
When an expert is being considered, that is a professional decision, not a software one. In civil proceedings in England and Wales, expert evidence is restricted to that which is reasonably required to resolve the proceedings, under Part 35 of the Civil Procedure Rules. The Civil Justice Council’s Guidance for the Instruction of Experts in Civil Claims sets out good practice for those who instruct experts. Even where a claim never reaches proceedings, use the organised file to see earlier whether a medical question is worth an expert’s time, then consult a person qualified to answer it.
The claim still needs a person
Organising the medical file sooner is a practical gain for insurers. Prior conditions and treatment gaps are less likely to arrive as a surprise if the production has been put in treatment order and each chronology entry can be checked against a page. That is work AI can take on.
What it cannot do is decide what those facts mean, whether the production is complete, whether the incident caused the injury, or whether an independent medical expert should be asked to advise. Those remain judgements for the claims handler and, where appropriate, for counsel and for the expert. The software can get the file into a shape in which those judgements can be made earlier. It should not be asked to make them.
About the author
Phil Worthey works in business development at Litegy, an AI medical record analysis platform for personal injury, clinical negligence, insurance-defence and claims teams.

Be the first to comment